Suggestion: Spatio-temporal analysis of outbreak - https://covid19.geonav.xyz/

[aleksandar-jovicic]

This is more suggestion for new web site features than issue with data.

I developed following site https://covid19.geonav.xyz/ that use your datasets but as update is done once a day it is not near-realtime ... so you can consider to add following features in your existing web app:

1. Temporal scale (showing situation at selected point in time) ... e.g. top slider on site from example
2. Showing epi-curves for selected cluster (or at least province) ... e.g. update side charts when cluster is selected
3. Introduction of case-fatality rate curve ... e.g. 3rd curve on right-hand panel from example
4. Someone already mentioned fatalities count on all epi-curves (can be challenging due to magnitude of difference between new cases and fatalities ... I'm thinking to separate those two into different charts)
5. Selection of color schema ... e.g. dark/bright map source on bottom-left of example
6. Heat-map and dynamic clustering can be challenging due to algorithms behind platform (I'm still playing with options but some more "intelligent" clustering will require different approach than provided by off-the-shelf solution of mapping library

[avatorl]

introduction of case-fatality rate curve ... e.g. 3rd curve on right-hand panel from example naive CFR calculated as deaths/cases is mathematically incorrect and misleading value http://avatorl.org/en/novel-coronavirus-2019-ncov-fatality-rate-who-and-media-vs-reality/

[bjarkimg]

Very nice viz. Agree on this deaths/cases, difficult to tell the true fatality rate at this point and may be misleading.

[bjarkimg]

@avatorl your power bi link broken? http://avatorl.org/covid-19/

[dawenx]

@bjarkimg power bi link works for me when I tested it just now

[bjarkimg]

Chrome... works in IE fine :) thanks.

[avatorl]

@bjarkimg Do you have a problem with the report in Chrome? What kind of problem?

[bjarkimg]

@avatorl Very nice analysis you have in this Power BI report and your blog. I will need to step up my game :) My PBI version -> http://bit.ly/38qeMbz

[bjarkimg]

@bjarkimg Do you have a problem with the report in Chrome? What kind of problem? see only this

[aleksandar-jovicic]

@avatorl , @bjarkimg ... as always such ratios are sensitive on small numbers ... and I agree that we cannot determine true CFR for disease at this moment (without serological tests we cannot be sure how much mild cases are not detected) but CFR curve from example can point to some issues in health system or patient management. If naive calculated CFR is significantly higher in some area than in another (off course paying attention that number of cases and fatalities are high enough) then it can be used for further investigation of causes (e.g. if some clinic is overloaded by patients ... or lack of medications ... or whatever can cause higher CFR comparing to others)

So CFR here is tool for issue detecting, not as scientific determination of CFR for disease in general.

[dawenx]

@avatorl i mentioned you the other day on this. Trying again since you are online now. :)

do you want to add other’s link in the tool? Mine is here: http://ncov.bii.virginia.edu/dashboard/

[avatorl]

@aleksandar-jovicic

So CFR here is tool for issue detecting

it will be different depends on how fast an epidemic is growing first of all. so it's useless and misleading if you want to detect issues.

@bjarkimg thanks thanks for the screenshot. it works a bit strange when embedded into inframe, but so far there was a problem only with missing bubbles on the maps (for some users only) and I've contacted Microsoft support regarding this problem. strange that it is totally blank for you.

[avatorl]

@dawenx thanks, I like your dashboard, but I'm not interested in adding links to other tools

[aleksandar-jovicic]

@avatorl ... sharp increase in number of fatalities usually happens when system is overloaded. With time series of CFR you can at least identify when it happens ... and with spatial clustering you can detect it on neighbour sources (not necessary only by one administratively separate). For sure it will be more benefits from fine granular data than this one (province statistics are too coarse). Off course ... as time passing CFR curve will be more and more smooth but even then one can detect spikes (and such spikes can be more important at later phases of outbreak than on early one)

[bjarkimg]

@avatorl Fixed now had to clear cache in Chrome for some strange reason... main PBI service was not working also..